Dataset of road surface images with seasons for machine learning applications

Road surface monitoring plays a vital role in ensuring safety and comfort for the various road users, from pedestrians to drivers. Furthermore, this information is useful for the maintenance of the roads. The road condition deteriorates due to volatile weather. Thus the main objective of the proposed paper is to create an image dataset of the road surface for two seasons, i.e. summer and rainy. Accordingly, we created road surface images for different roads such as paved and unpaved roads. These folders consist of two subfolders for Rainy and Summer potholes. The dataset consists of 8484 images and 10 videos. This dataset is highly useful for machine learning experts working in the field of automatic vehicle controlling and road surface monitoring.


Specifications
Computer Vision and Pattern Recognition Specific subject area Road Surface Detection Type of data Image, Video How datapoints were acquired

Value of the Data
• This dataset is essential as it contributes to the future applications of the Sustainable Development Goal-11 of the United Nations, i.e., "sustainable-transport'' (UN SDG -11) [3] . We concentrated on the forms of road surface available during different seasons to achieve accuracy in road surface detection, which makes this dataset unique. • These datapoints are available in the public repository for all the Research Institutes, Scientific communities and Policymakers. • This is the first dataset that provides road surface images according to seasons that could be used in a wide variety of future research studies related to road accident prevention and surface inspection [4] . • The data may be reused for conducting experiments related to road accident prevention, Pothole detection, road surface water detection, automatic vehicle controlling mechanism and path navigation mechanism. Moreover, the researchers involved in earth surface inspection may benefit indirectly [2] .  sidered, such as (a) Pothole in rains on the paved and unpaved road, (b) Pothole in summer on the paved and unpaved road (c) Speed breaker and (d) Roadside Barriers. Cement concrete and tar roads were considered for the paved road condition, while roads made of only soil and stones were unpaved. All the images were taken from the top view and the side view. These images were captured during 2020 and 2021 from two cities, namely Nashik and Mumbai, of Maharashtra, India.

Data Description
Based on the road structure, the directory structure is divided into three main folders: (1) Paved Road, (2) Unpaved and (3) Video. The first two folders are further divided into subcategories according to the status of the pothole during summer and rains. Along with potholes, we also added speed breaker and roadside barrier images. The video folder consists of videos according to seasons for various road surface types ( Fig. 2 ).

Experimental design
The image data acquisition process is shown in Fig. 3 . The Road surface images were acquired using the Samsung Galaxy A22 mobile's high-resolution quad rear camera. In all, 8484 images were captured using a camera and then were segregated and saved in respective folders as per the Road type and class of the image ( Fig. 2 ). The 10 videos were captured for various road surfaces and stored in the Video folder.
The dataset is collected for the images available in two seasons, i.e. Rainy and Summer. Therefore, the duration from April to December is chosen for capturing images ( Table 1 ).
The road surface was available in different forms such as speed breakers, potholes and roadside barriers ( Fig. 1 ). Thus, the dataset was prepared by considering all adverse conditions required for Road surface detection according to season. To cope with adverse conditions, it is necessary to collect images in different illuminations. Hence the images were captured in the morning, afternoon, evening, and late evening from the top and the side view ( Table 4 ).

Materials or specification of image acquisition system
The road images were captured using the Samsung Galaxy A22 RGB Quad camera ( Table 2 ) [5] . A 15 W battery was used to power all the components of the imaging system. All dataset   Table 4 describes the classes, number of images taken and the environments in which images are taken. A handheld mobile camera was used to capture images from the top view and side view. The images were captured at a man's height by bending down. The images of speed  breakers, roadside barriers and potholes during different seasons for paved and unpaved roads were included.

Ethics Statement
This data is available in the public domain, and no funding is received for the present effort. There is no conflict of interest.